Update CPU & RAM install requirements
Problem to solve
The documented install requirements, especially around CPU and RAM requirements, are confusing and potentially misleading.
Further details
Here are the problems that I see:
4 cores supports up to 2,000 users
Saying X number of cores supports up to Y number of users seems misleading. This figure might be close in some cases, but I think in many cases this is inaccurate. Scaling a system is too complex of a matter to simplify down to this level. The certainty of the wording is especially concerning, because as a user I may expect to confidently support 2000 users on a system with 4 cores and be surprised when it doesn't work for me.
I'd rather see minimum, recommended, and a scaling guide (probably linked somewhere else). The "minimum" should be the minimum that we recommend for setting up a GitLab instance to start experimenting with, but not use in production. The "recommended" is what you can reasonably get away with to begin running a production GitLab system. The scaling guide will attempt to outline how to scale (horizontally or vertically) when the "recommended" specs aren't enough for the user's workload.
2 cores is the recommended number of cores and supports up to 500 users
Is 2 really the recommended number of cores? Maybe we could say that 2 is the recommended minimum and that 4 is the recommended number of cores.
More users? Run it on multiple application servers
We have this blurb in both the RAM section and the CPU section. We seem to be implying that you should only look at horizontal scaling after reaching somewhere in the ballpark 30k to 40k users. I don't this is accurate or fair.
Proposal
I think these sections can be redone and scaling info can be moved into a "scaling gitlab" guide. I think the specs could be simplified to something like the following:
CPU
2 cores is the minimum that you should use with GitLab. We recommend at least 4 cores for a production GitLab system.
8GB of ram is the minimum that you should use to run GitLab. This may work on a production system, but more is better.
Who can address the issue
Somebody with more experience in scaling GitLab systems. Maybe somebody on our infrastructure team.